1 /*************************************************
2 * Exim - an Internet mail transport agent *
3 *************************************************/
5 /* Copyright (c) University of Cambridge 1995 - 2018 */
6 /* Copyright (c) The Exim Maintainers 2020 - 2021 */
7 /* See the file NOTICE for conditions of use and distribution. */
9 /* Functions for matching strings */
15 /* Argument block for the check_string() function. This is used for general
16 strings, domains, and local parts. */
18 typedef struct check_string_block {
19 const uschar *origsubject; /* caseful; keep these two first, in */
20 const uschar *subject; /* step with the block below */
28 /* Argument block for the check_address() function. This is used for whole
31 typedef struct check_address_block {
32 const uschar *origaddress; /* caseful; keep these two first, in */
33 uschar *address; /* step with the block above */
36 } check_address_block;
40 /*************************************************
41 * Generalized string match *
42 *************************************************/
44 /* This function does a single match of a subject against a pattern, and
45 optionally sets up the numeric variables according to what it matched. It is
46 called from match_isinlist() via match_check_list() when scanning a list, and
47 from match_check_string() when testing just a single item. The subject and
48 options arguments are passed in a check_string_block so as to make it easier to
49 pass them through match_check_list.
51 The possible types of pattern are:
53 . regular expression - starts with ^
54 . tail match - starts with *
55 . lookup - starts with search type
56 . if at_is_special is set in the argument block:
57 @ matches the primary host name
58 @[] matches a local IP address in brackets
59 @mx_any matches any domain with an MX to the local host
60 @mx_primary matches any domain with a primary MX to the local host
61 @mx_secondary matches any domain with a secondary MX to the local host
62 . literal - anything else
64 Any of the @mx_xxx options can be followed by "/ignore=<list>" where <list> is
65 a list of IP addresses that are to be ignored (typically 127.0.0.1).
68 arg check_string_block pointer - see below
69 pattern the pattern to be matched
70 valueptr if not NULL, and a lookup is done, return the result here
71 instead of discarding it; else set it to point to NULL
72 error for error messages (not used in this function; it never
75 Contents of the argument block:
76 origsubject the subject in its original casing
77 subject the subject string to be checked, lowercased if caseless
78 expand_setup if < 0, don't set up any numeric expansion variables;
79 if = 0, set $0 to whole subject, and either
80 $1 to what matches * or
81 $1, $2, ... to r.e. bracketed items
82 if > 0, don't set $0, but do set either
83 $n to what matches *, or
84 $n, $n+1, ... to r.e. bracketed items
85 (where n = expand_setup)
86 use_partial if FALSE, override any partial- search types
87 caseless TRUE for caseless matching where possible
88 at_is_special enable special handling of items starting with @
90 Returns: OK if matched
92 DEFER if lookup deferred
96 check_string(void *arg, const uschar *pattern, const uschar **valueptr, uschar **error)
98 const check_string_block *cb = arg;
99 int search_type, partial, affixlen, starflags;
100 int expand_setup = cb->expand_setup;
101 const uschar * affix, * opts;
103 uschar *filename = NULL;
104 uschar *keyquery, *result, *semicolon;
107 if (valueptr) *valueptr = NULL;
109 /* For regular expressions, use cb->origsubject rather than cb->subject so that
110 it works if the pattern uses (?-i) to turn off case-independence, overriding
113 s = string_copy(pattern[0] == '^' ? cb->origsubject : cb->subject);
115 /* If required to set up $0, initialize the data but don't turn on by setting
116 expand_nmax until the match is assured. */
119 if (expand_setup == 0)
121 expand_nstring[0] = s; /* $0 (might be) the matched subject in full */
122 expand_nlength[0] = Ustrlen(s);
124 else if (expand_setup > 0) expand_setup--;
126 /* Regular expression match: compile, match, and set up $ variables if
129 if (pattern[0] == '^')
131 const pcre2_code * re = regex_must_compile(pattern, cb->caseless, FALSE);
133 ? !regex_match(re, s, -1, NULL)
134 : !regex_match_and_setup(re, s, 0, expand_setup)
137 if (valueptr) *valueptr = pattern; /* "value" gets the RE */
143 if (pattern[0] == '*')
145 int slen = Ustrlen(s);
146 int patlen; /* Sun compiler doesn't like non-constant initializer */
148 patlen = Ustrlen(++pattern);
149 if (patlen > slen) return FAIL;
151 ? strncmpic(s + slen - patlen, pattern, patlen) != 0
152 : Ustrncmp(s + slen - patlen, pattern, patlen) != 0)
154 if (expand_setup >= 0)
156 expand_nstring[++expand_setup] = s; /* write a $n, the matched subject variable-part */
157 expand_nlength[expand_setup] = slen - patlen;
158 expand_nmax = expand_setup; /* commit also $0, the matched subject */
160 if (valueptr) *valueptr = pattern - 1; /* "value" gets the (original) pattern */
164 /* Match a special item starting with @ if so enabled. On its own, "@" matches
165 the primary host name - implement this by changing the pattern. For the other
166 cases we have to do some more work. If we don't recognize a special pattern,
167 just fall through - the match will fail. */
169 if (cb->at_is_special && pattern[0] == '@')
173 pattern = primary_hostname;
174 goto NOT_AT_SPECIAL; /* Handle as exact string match */
177 if (Ustrcmp(pattern, "@[]") == 0)
179 int slen = Ustrlen(s);
180 if (s[0] != '[' && s[slen-1] != ']') return FAIL; /*XXX should this be || ? */
181 for (ip_address_item * ip = host_find_interfaces(); ip; ip = ip->next)
182 if (Ustrncmp(ip->address, s+1, slen - 2) == 0
183 && ip->address[slen - 2] == 0)
185 if (expand_setup >= 0) expand_nmax = expand_setup; /* commit $0, the IP addr */
186 if (valueptr) *valueptr = pattern; /* "value" gets the pattern */
192 if (strncmpic(pattern, US"@mx_", 4) == 0)
198 BOOL removed = FALSE;
199 const uschar *ss = pattern + 4;
200 const uschar *ignore_target_hosts = NULL;
202 if (strncmpic(ss, US"any", 3) == 0) ss += 3;
203 else if (strncmpic(ss, US"primary", 7) == 0)
208 else if (strncmpic(ss, US"secondary", 9) == 0)
213 else goto NOT_AT_SPECIAL;
215 if (strncmpic(ss, US"/ignore=", 8) == 0) ignore_target_hosts = ss + 8;
216 else if (*ss) goto NOT_AT_SPECIAL;
222 rc = host_find_bydns(&h,
224 HOST_FIND_BY_MX, /* search only for MX, not SRV or A */
225 NULL, /* service name not relevant */
226 NULL, /* srv_fail_domains not relevant */
227 NULL, /* mx_fail_domains not relevant */
228 NULL, /* no dnssec request/require XXX ? */
229 NULL, /* no feedback FQDN */
230 &removed); /* feedback if local removed */
232 if (rc == HOST_FIND_AGAIN)
234 search_error_message = string_sprintf("DNS lookup of \"%s\" deferred", s);
238 if ((rc != HOST_FOUND_LOCAL || secy) && (prim || !removed))
241 if (expand_setup >= 0) expand_nmax = expand_setup; /* commit $0, the matched subject */
242 if (valueptr) *valueptr = pattern; /* "value" gets the patterm */
245 /*** The above line used to be the following line, but this is incorrect,
246 because host_find_bydns() may return HOST_NOT_FOUND if it removed some MX
247 hosts, but the remaining ones were non-existent. All we are interested in
248 is whether or not it removed some hosts.
250 return (rc == HOST_FOUND && removed)? OK : FAIL;
255 /* Escape point from code for specials that start with "@" */
259 /* This is an exact string match if there is no semicolon in the pattern. */
261 if ((semicolon = Ustrchr(pattern, ';')) == NULL)
263 if (cb->caseless ? strcmpic(s, pattern) != 0 : Ustrcmp(s, pattern) != 0)
265 if (expand_setup >= 0) expand_nmax = expand_setup; /* Original code! $0 gets the matched subject */
266 if (valueptr) *valueptr = pattern; /* "value" gets the pattern */
270 /* Otherwise we have a lookup item. The lookup type, including partial, etc. is
271 the part of the string preceding the semicolon. */
274 search_type = search_findtype_partial(pattern, &partial, &affix, &affixlen,
277 if (search_type < 0) log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
278 search_error_message);
280 /* Partial matching is not appropriate for certain lookups (e.g. when looking
281 up user@domain for sender rejection). There's a flag to disable it. */
283 if (!cb->use_partial) partial = -1;
285 /* Set the parameters for the three different kinds of lookup. */
287 keyquery = search_args(search_type, s, semicolon+1, &filename, opts);
289 /* Now do the actual lookup; throw away the data returned unless it was asked
290 for; partial matching is all handled inside search_find(). Note that there is
291 no search_close() because of the caching arrangements. */
293 if (!(handle = search_open(filename, search_type, 0, NULL, NULL)))
294 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s", search_error_message);
295 result = search_find(handle, filename, keyquery, partial, affix, affixlen,
296 starflags, &expand_setup, opts);
298 if (!result) return f.search_find_defer ? DEFER : FAIL;
299 if (valueptr) *valueptr = result;
301 expand_nmax = expand_setup;
307 /*************************************************
308 * Public interface to check_string() *
309 *************************************************/
311 /* This function is called from several places where is it most convenient to
312 pass the arguments individually. It places them in a check_string_block
313 structure, and then calls check_string().
316 s the subject string to be checked
317 pattern the pattern to check it against
318 expand_setup expansion setup option (see check_string())
319 use_partial if FALSE, override any partial- search types
320 caseless TRUE for caseless matching where possible
321 at_is_special TRUE to recognize @, @[], etc.
322 valueptr if not NULL, and a file lookup was done, return the result
323 here instead of discarding it; else set it to point to NULL
325 Returns: OK if matched
327 DEFER if lookup deferred
331 match_check_string(const uschar *s, const uschar *pattern, int expand_setup,
332 BOOL use_partial, BOOL caseless, BOOL at_is_special, const uschar **valueptr)
334 check_string_block cb;
336 cb.subject = caseless ? string_copylc(s) : string_copy(s);
337 cb.expand_setup = expand_setup;
338 cb.use_partial = use_partial;
339 cb.caseless = caseless;
340 cb.at_is_special = at_is_special;
341 return check_string(&cb, pattern, valueptr, NULL);
346 /*************************************************
347 * Get key string from check block *
348 *************************************************/
350 /* When caching the data from a lookup for a named list, we have to save the
351 key that was found, because other lookups of different keys on the same list
352 may occur. This function has knowledge of the different lookup types, and
353 extracts the appropriate key.
357 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
360 static const uschar *
361 get_check_key(void *arg, int type)
368 return ((check_string_block *)arg)->subject;
371 return ((check_host_block *)arg)->host_address;
374 return ((check_address_block *)arg)->address;
376 return US""; /* In practice, should never happen */
381 /*************************************************
382 * Scan list and run matching function *
383 *************************************************/
385 /* This function scans a list of patterns, and runs a matching function for
386 each item in the list. It is called from the functions that match domains,
387 local parts, hosts, and addresses, because its overall structure is the same in
388 all cases. However, the details of each particular match is different, so it
389 calls back to a given function do perform an actual match.
391 We can't quite keep the different types anonymous here because they permit
392 different special cases. A pity.
394 If a list item starts with !, that implies negation if the subject matches the
395 rest of the item (ignoring white space after the !). The result when the end of
396 the list is reached is FALSE unless the last item on the list is negated, in
397 which case it is TRUE. A file name in the list causes its lines to be
398 interpolated as if items in the list. An item starting with + is a named
399 sublist, obtained by searching the tree pointed to by anchorptr, with possible
400 cached match results in cache_bits.
403 listptr pointer to the pointer to the list
404 sep separator character for string_nextinlist();
405 normally zero for a standard list;
406 sometimes UCHAR_MAX+1 for single items;
407 anchorptr -> tree of named items, or NULL if no named items
408 cache_ptr pointer to pointer to cache bits for named items, or
409 pointer to NULL if not caching; may get set NULL if an
410 uncacheable named list is encountered
411 func function to call back to do one test
412 arg pointer to pass to the function; the string to be matched is
413 in the structure it points to
414 type MCL_STRING, MCL_DOMAIN, MCL_HOST, MCL_ADDRESS, or MCL_LOCALPART
415 these are used for some special handling
416 MCL_NOEXPAND (whose value is greater than any of them) may
417 be added to any value to suppress expansion of the list
418 name string to use in debugging info
419 valueptr where to pass back data from a lookup
421 Returns: OK if matched a non-negated item
422 OK if hit end of list after a negated item
423 FAIL if expansion force-failed
424 FAIL if matched a negated item
425 FAIL if hit end of list after a non-negated item
426 DEFER if a something deferred or expansion failed
430 match_check_list(const uschar **listptr, int sep, tree_node **anchorptr,
431 unsigned int **cache_ptr, int (*func)(void *,const uschar *,const uschar **,uschar **),
432 void *arg, int type, const uschar *name, const uschar **valueptr)
435 unsigned int * original_cache_bits = *cache_ptr;
436 BOOL include_unknown = FALSE, ignore_unknown = FALSE,
437 include_defer = FALSE, ignore_defer = FALSE;
442 /* Save time by not scanning for the option name when we don't need it. */
446 uschar * listname = readconf_find_option(listptr);
447 if (*listname) ot = string_sprintf("%s in %s?", name, listname);
450 /* If the list is empty, the answer is no. Skip the debugging output for
455 HDEBUG(D_lists) if (ot) debug_printf_indent("%s no (option unset)\n", ot);
459 /* Expand the list before we scan it. A forced expansion gives the answer
460 "not in list"; other expansion errors cause DEFER to be returned. However,
461 if the type value is greater than or equal to than MCL_NOEXPAND, do not expand
464 if (type >= MCL_NOEXPAND)
467 type -= MCL_NOEXPAND; /* Remove the "no expand" flag */
471 /* If we are searching a domain list, and $domain is not set, set it to the
472 subject that is being sought for the duration of the expansion. */
474 if (type == MCL_DOMAIN && !deliver_domain)
476 check_string_block *cb = (check_string_block *)arg;
477 deliver_domain = string_copy(cb->subject);
478 list = expand_cstring(*listptr);
479 deliver_domain = NULL;
482 list = expand_cstring(*listptr);
486 if (f.expand_string_forcedfail)
488 HDEBUG(D_lists) debug_printf_indent("expansion of \"%s\" forced failure: "
489 "assume not in this list\n", *listptr);
492 log_write(0, LOG_MAIN|LOG_PANIC, "failed to expand \"%s\" while checking "
493 "a list: %s", *listptr, expand_string_message);
498 /* For an unnamed list, use the expanded version in comments */
499 #define LIST_LIMIT_PR 2048
501 HDEBUG(D_any) if (!ot)
504 gstring * g = string_fmt_append(NULL, "%s in \"%n%.*s%n\"",
505 name, &n, LIST_LIMIT_PR, list, &m);
506 if (m - n >= LIST_LIMIT_PR) g = string_catn(g, US"...", 3);
507 g = string_catn(g, US"?", 1);
508 gstring_release_unused(g);
509 ot = string_from_gstring(g);
512 /* Now scan the list and process each item in turn, until one of them matches,
513 or we hit an error. */
515 while ((sss = string_nextinlist(&list, &sep, NULL, 0)))
519 /* Address lists may contain +caseful, to restore caseful matching of the
520 local part. We have to know the layout of the control block, unfortunately.
521 The lower cased address is in a temporary buffer, so we just copy the local
522 part back to the start of it (if a local part exists). */
524 if (type == MCL_ADDRESS)
526 if (Ustrcmp(ss, "+caseful") == 0)
528 check_address_block *cb = (check_address_block *)arg;
529 uschar *at = Ustrrchr(cb->origaddress, '@');
532 Ustrncpy(cb->address, cb->origaddress, at - cb->origaddress);
533 cb->caseless = FALSE;
538 /* Similar processing for local parts */
540 else if (type == MCL_LOCALPART)
542 if (Ustrcmp(ss, "+caseful") == 0)
544 check_string_block *cb = (check_string_block *)arg;
545 Ustrcpy(US cb->subject, cb->origsubject);
546 cb->caseless = FALSE;
551 /* If the host item is "+include_unknown" or "+ignore_unknown", remember it
552 in case there's a subsequent failed reverse lookup. There is similar
553 processing for "defer". */
555 else if (type == MCL_HOST && *ss == '+')
557 if (Ustrcmp(ss, "+include_unknown") == 0)
559 include_unknown = TRUE;
560 ignore_unknown = FALSE;
563 if (Ustrcmp(ss, "+ignore_unknown") == 0)
565 ignore_unknown = TRUE;
566 include_unknown = FALSE;
569 if (Ustrcmp(ss, "+include_defer") == 0)
571 include_defer = TRUE;
572 ignore_defer = FALSE;
575 if (Ustrcmp(ss, "+ignore_defer") == 0)
578 include_defer = FALSE;
583 /* Starting with ! specifies a negative item. It is theoretically possible
584 for a local part to start with !. In that case, a regex has to be used. */
589 while (isspace((*(++ss))));
594 /* If the item does not begin with '/', it might be a + item for a named
595 list. Otherwise, it is just a single list entry that has to be matched.
596 We recognize '+' only when supplied with a tree of named lists. */
600 if (*ss == '+' && anchorptr)
605 unsigned int *use_cache_bits = original_cache_bits;
606 uschar *cached = US"";
610 if (!(t = tree_search(*anchorptr, ss+1)))
612 log_write(0, LOG_MAIN|LOG_PANIC, "unknown named%s list \"%s\"",
613 type == MCL_DOMAIN ? " domain" :
614 type == MCL_HOST ? " host" :
615 type == MCL_ADDRESS ? " address" :
616 type == MCL_LOCALPART ? " local part" : "",
622 /* If the list number is negative, it means that this list is not
623 cacheable because it contains expansion items. */
625 if (nb->number < 0) use_cache_bits = NULL;
627 /* If we have got a cache pointer, get the bits. This is not an "else"
628 because the pointer may be NULL from the start if caching is not
633 offset = (nb->number)/16;
634 shift = ((nb->number)%16)*2;
635 bits = use_cache_bits[offset] & (3 << shift);
638 /* Not previously tested or no cache - run the full test */
642 switch (match_check_list(&(nb->string), 0, anchorptr, &use_cache_bits,
643 func, arg, type, name, valueptr))
645 case OK: bits = 1; break;
646 case FAIL: bits = 3; break;
647 case DEFER: goto DEFER_RETURN;
650 /* If this list was uncacheable, or a sublist turned out to be
651 uncacheable, the value of use_cache_bits will now be NULL, even if it
652 wasn't before. Ensure that this is passed up to the next level.
653 Otherwise, remember the result of the search in the cache. */
659 use_cache_bits[offset] |= bits << shift;
663 int old_pool = store_pool;
664 namedlist_cacheblock *p;
666 /* Cached data for hosts persists over more than one message,
667 so we use the permanent store pool */
669 store_pool = POOL_PERM;
670 p = store_get(sizeof(namedlist_cacheblock), GET_UNTAINTED);
671 p->key = string_copy(get_check_key(arg, type));
674 p->data = *valueptr ? string_copy(*valueptr) : NULL;
675 store_pool = old_pool;
677 p->next = nb->cache_data;
680 DEBUG(D_lists) debug_printf_indent("data from lookup saved for "
681 "cache for %s: key '%s' value '%s'\n", ss, p->key, *valueptr);
686 /* Previously cached; to find a lookup value, search a chain of values
687 and compare keys. Typically, there is only one such, but it is possible
688 for different keys to have matched the same named list. */
692 DEBUG(D_lists) debug_printf_indent("cached %s match for %s\n",
693 (bits & (-bits)) == bits ? "yes" : "no", ss);
695 cached = US" - cached";
698 const uschar *key = get_check_key(arg, type);
700 for (namedlist_cacheblock * p = nb->cache_data; p; p = p->next)
701 if (Ustrcmp(key, p->key) == 0)
706 DEBUG(D_lists) debug_printf_indent("cached lookup data = %s\n", *valueptr);
710 /* Result of test is indicated by value in bits. For each test, we
711 have 00 => untested, 01 => tested yes, 11 => tested no. */
713 if ((bits & (-bits)) == bits) /* Only one of the two bits is set */
715 HDEBUG(D_lists) debug_printf_indent("%s %s (matched \"%s\"%s)\n", ot,
716 yield == OK ? "yes" : "no", sss, cached);
721 /* Run the provided function to do the individual test. */
725 uschar * error = NULL;
726 switch ((func)(arg, ss, valueptr, &error))
729 HDEBUG(D_lists) debug_printf_indent("%s %s (matched \"%s\")\n", ot,
730 (yield == OK)? "yes" : "no", sss);
735 error = string_sprintf("DNS lookup of \"%s\" deferred", ss);
738 HDEBUG(D_lists) debug_printf_indent("%s: item ignored by +ignore_defer\n",
744 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
747 if (!search_error_message) search_error_message = error;
750 /* The ERROR return occurs when checking hosts, when either a forward
751 or reverse lookup has failed. It can also occur in a match_ip list if a
752 non-IP address item is encountered. The error string gives details of
758 HDEBUG(D_lists) debug_printf_indent("%s: item ignored by +ignore_unknown\n",
763 HDEBUG(D_lists) debug_printf_indent("%s %s (%s)\n", ot,
764 include_unknown? "yes":"no", error);
765 if (!include_unknown)
767 if (LOGGING(unknown_in_list))
768 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
771 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
778 /* If the item is a file name, we read the file and do a match attempt
779 on each line in the file, including possibly more negation processing. */
783 int file_yield = yield; /* In case empty file */
784 uschar * filename = ss;
785 FILE * f = Ufopen(filename, "rb");
786 uschar filebuffer[1024];
788 /* ot will be null in non-debugging cases, and anyway, we get better
789 wording by reworking it. */
793 uschar * listname = readconf_find_option(listptr);
794 if (listname[0] == 0)
795 listname = string_sprintf("\"%s\"", *listptr);
796 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "%s",
797 string_open_failed("%s when checking %s", sss, listname));
800 /* Trailing comments are introduced by #, but in an address list or local
801 part list, the # must be preceded by white space or the start of the line,
802 because the # character is a legal character in local parts. */
804 while (Ufgets(filebuffer, sizeof(filebuffer), f) != NULL)
807 uschar *sss = filebuffer;
809 while ((ss = Ustrchr(sss, '#')) != NULL)
811 if ((type != MCL_ADDRESS && type != MCL_LOCALPART) ||
812 ss == filebuffer || isspace(ss[-1]))
820 ss = filebuffer + Ustrlen(filebuffer); /* trailing space */
821 while (ss > filebuffer && isspace(ss[-1])) ss--;
825 while (isspace(*ss)) ss++; /* leading space */
827 if (!*ss) continue; /* ignore empty */
829 file_yield = yield; /* positive yield */
830 sss = ss; /* for debugging */
832 if (*ss == '!') /* negation */
834 file_yield = (file_yield == OK)? FAIL : OK;
835 while (isspace((*(++ss))));
838 switch ((func)(arg, ss, valueptr, &error))
842 HDEBUG(D_lists) debug_printf_indent("%s %s (matched \"%s\" in %s)\n", ot,
843 yield == OK ? "yes" : "no", sss, filename);
845 /* The "pattern" being matched came from the file; we use a stack-local.
846 Copy it to allocated memory now we know it matched. */
848 if (valueptr) *valueptr = string_copy(ss);
853 error = string_sprintf("DNS lookup of %s deferred", ss);
856 HDEBUG(D_lists) debug_printf_indent("%s: item ignored by +ignore_defer\n",
863 log_write(0, LOG_MAIN, "%s: accepted by +include_defer", error);
868 case ERROR: /* host name lookup failed - this can only */
869 if (ignore_unknown) /* be for an incoming host (not outgoing) */
871 HDEBUG(D_lists) debug_printf_indent("%s: item ignored by +ignore_unknown\n",
876 HDEBUG(D_lists) debug_printf_indent("%s %s (%s)\n", ot,
877 include_unknown? "yes":"no", error);
879 if (!include_unknown)
881 if (LOGGING(unknown_in_list))
882 log_write(0, LOG_MAIN, "list matching forced to fail: %s", error);
885 log_write(0, LOG_MAIN, "%s: accepted by +include_unknown", error);
891 /* At the end of the file, leave the yield setting at the final setting
892 for the file, in case this is the last item in the list. */
897 } /* Loop for the next item on the top-level list */
899 /* End of list reached: if the last item was negated yield OK, else FAIL. */
902 debug_printf_indent("%s %s (end of list)\n", ot, yield == OK ? "no":"yes");
903 return yield == OK ? FAIL : OK;
905 /* Something deferred */
908 HDEBUG(D_lists) debug_printf("%s list match deferred for %s\n", ot, sss);
913 /*************************************************
914 * Match in colon-separated list *
915 *************************************************/
917 /* This function is used for domain lists and local part lists. It is not used
918 for host lists or address lists, which have additional interpretation of the
919 patterns. Some calls of it set sep > UCHAR_MAX in order to use its matching
920 facilities on single items. When this is done, it arranges to set the numerical
921 variables as a result of the match.
923 This function is now just a short interface to match_check_list(), which does
924 list scanning in a general way. A good compiler will optimize the tail
928 s string to search for
929 listptr ptr to ptr to colon separated list of patterns, or NULL
930 sep a separator value for the list (see string_nextinlist())
931 anchorptr ptr to tree for named items, or NULL if no named items
932 cache_bits ptr to cache_bits for ditto, or NULL if not caching
933 type MCL_DOMAIN when matching a domain list
934 MCL_LOCALPART when matching a local part list (address lists
935 have their own function)
936 MCL_STRING for others (e.g. list of ciphers)
937 MCL_NOEXPAND (whose value is greater than any of them) may
938 be added to any value to suppress expansion of the list
939 caseless TRUE for (mostly) caseless matching - passed directly to
941 valueptr pointer to where any lookup data is to be passed back,
942 or NULL (just passed on to match_check_string)
944 Returns: OK if matched a non-negated item
945 OK if hit end of list after a negated item
946 FAIL if expansion force-failed
947 FAIL if matched a negated item
948 FAIL if hit end of list after a non-negated item
949 DEFER if a lookup deferred
953 match_isinlist(const uschar *s, const uschar **listptr, int sep,
954 tree_node **anchorptr,
955 unsigned int *cache_bits, int type, BOOL caseless, const uschar **valueptr)
957 unsigned int *local_cache_bits = cache_bits;
958 check_string_block cb;
960 cb.subject = caseless ? string_copylc(s) : string_copy(s);
961 cb.at_is_special = FALSE;
962 switch (type & ~MCL_NOEXPAND)
964 case MCL_DOMAIN: cb.at_is_special = TRUE; /*FALLTHROUGH*/
965 case MCL_LOCALPART: cb.expand_setup = 0; break;
966 default: cb.expand_setup = sep > UCHAR_MAX ? 0 : -1; break;
968 cb.use_partial = TRUE;
969 cb.caseless = caseless;
970 if (valueptr) *valueptr = NULL;
971 return match_check_list(listptr, sep, anchorptr, &local_cache_bits,
972 check_string, &cb, type, s, valueptr);
977 /*************************************************
978 * Match address to single address-list item *
979 *************************************************/
981 /* This function matches an address to an item from an address list. It is
982 called from match_address_list() via match_check_list(). That is why most of
983 its arguments are in an indirect block.
986 arg the argument block (see below)
987 pattern the pattern to match
988 valueptr where to return a value
989 error for error messages (not used in this function; it never
992 The argument block contains:
993 address the start of the subject address; when called from retry.c
994 it may be *@domain if the local part isn't relevant
995 origaddress the original, un-case-forced address (not used here, but used
996 in match_check_list() when +caseful is encountered)
997 expand_setup controls setting up of $n variables
998 caseless TRUE for caseless local part matching
1000 Returns: OK for a match
1002 DEFER if a lookup deferred
1006 check_address(void *arg, const uschar *pattern, const uschar **valueptr, uschar **error)
1008 check_address_block *cb = (check_address_block *)arg;
1009 check_string_block csb;
1012 unsigned int *null = NULL;
1013 const uschar *listptr;
1014 uschar *subject = cb->address;
1016 uschar *pdomain, *sdomain;
1018 DEBUG(D_lists) debug_printf_indent("address match test: subject=%s pattern=%s\n",
1021 /* Find the subject's domain */
1023 sdomain = Ustrrchr(subject, '@');
1025 /* The only case where a subject may not have a domain is if the subject is
1026 empty. Otherwise, a subject with no domain is a serious configuration error. */
1028 if (sdomain == NULL && *subject != 0)
1030 log_write(0, LOG_MAIN|LOG_PANIC, "no @ found in the subject of an "
1031 "address list match: subject=\"%s\" pattern=\"%s\"", subject, pattern);
1035 /* Handle a regular expression, which must match the entire incoming address.
1036 This may be the empty address. */
1038 if (*pattern == '^')
1039 return match_check_string(subject, pattern, cb->expand_setup, TRUE,
1040 cb->caseless, FALSE, NULL);
1042 /* Handle a pattern that is just a lookup. Skip over possible lookup names
1043 (letters, digits, hyphens). Skip over a possible * or *@ at the end. Then we
1044 must have a semicolon for it to be a lookup. */
1046 for (s = pattern; isalnum(*s) || *s == '-'; s++);
1050 /* If it is a straight lookup, do a lookup for the whole address. This may be
1051 the empty address. Partial matching doesn't make sense here, so we ignore it,
1052 but write a panic log entry. However, *@ matching will be honoured. */
1056 if (Ustrncmp(pattern, "partial-", 8) == 0)
1057 log_write(0, LOG_MAIN|LOG_PANIC, "partial matching is not applicable to "
1058 "whole-address lookups: ignored \"partial-\" in \"%s\"", pattern);
1059 return match_check_string(subject, pattern, -1, FALSE, cb->caseless, FALSE,
1063 /* For the remaining cases, an empty subject matches only an empty pattern,
1064 because other patterns expect to have a local part and a domain to match
1067 if (*subject == 0) return (*pattern == 0)? OK : FAIL;
1069 /* If the pattern starts with "@@" we have a split lookup, where the domain is
1070 looked up to obtain a list of local parts. If the subject's local part is just
1071 "*" (called from retry) the match always fails. */
1073 if (pattern[0] == '@' && pattern[1] == '@')
1078 if (sdomain == subject + 1 && *subject == '*') return FAIL;
1080 /* Loop for handling chains. The last item in any list may be of the form
1081 ">name" in order to chain on to another list. */
1083 for (const uschar * key = sdomain + 1; key && watchdog-- > 0; )
1087 if ((rc = match_check_string(key, pattern + 2, -1, TRUE, FALSE, FALSE,
1088 CUSS &list)) != OK) return rc;
1090 /* Check for chaining from the last item; set up the next key if one
1093 ss = Ustrrchr(list, ':');
1094 if (ss == NULL) ss = list; else ss++;
1095 while (isspace(*ss)) ss++;
1099 while (isspace(*ss)) ss++;
1100 key = string_copy(ss);
1104 /* Look up the local parts provided by the list; negation is permitted.
1105 If a local part has to begin with !, a regex can be used. */
1107 while ((ss = string_nextinlist(CUSS &list, &sep, NULL, 0)))
1114 while (isspace((*(++ss))));
1116 else local_yield = OK;
1119 rc = match_check_string(subject, ss, -1, TRUE, cb->caseless, FALSE,
1134 /* End of chain loop; panic if too many times */
1137 log_write(0, LOG_MAIN|LOG_PANIC_DIE, "Loop detected in lookup of "
1138 "local part of %s in %s", subject, pattern);
1140 /* Otherwise the local part check has failed, so the whole match
1147 /* We get here if the pattern is not a lookup or a regular expression. If it
1148 contains an @ there is both a local part and a domain. */
1150 pdomain = Ustrrchr(pattern, '@');
1151 if (pdomain != NULL)
1155 /* If the domain in the pattern is empty or one of the special cases [] or
1156 mx_{any,primary,secondary}, and the local part in the pattern ends in "@",
1157 we have a pattern of the form <something>@@, <something>@@[], or
1158 <something>@@mx_{any,primary,secondary}. These magic "domains" are
1159 automatically interpreted in match_check_string. We just need to arrange that
1160 the leading @ is included in the domain. */
1162 if (pdomain > pattern && pdomain[-1] == '@' &&
1164 Ustrcmp(pdomain+1, "[]") == 0 ||
1165 Ustrcmp(pdomain+1, "mx_any") == 0 ||
1166 Ustrcmp(pdomain+1, "mx_primary") == 0 ||
1167 Ustrcmp(pdomain+1, "mx_secondary") == 0))
1170 pllen = pdomain - pattern;
1171 sllen = sdomain - subject;
1173 /* Compare the local parts in the subject and the pattern */
1175 if (*pattern == '*')
1177 int cllen = pllen - 1;
1178 if (sllen < cllen) return FAIL;
1180 ? strncmpic(subject+sllen-cllen, pattern + 1, cllen) != 0
1181 : Ustrncmp(subject+sllen-cllen, pattern + 1, cllen) != 0)
1183 if (cb->expand_setup > 0)
1185 expand_nstring[cb->expand_setup] = subject;
1186 expand_nlength[cb->expand_setup] = sllen - cllen;
1192 if (sllen != pllen) return FAIL;
1194 ? strncmpic(subject, pattern, sllen) != 0
1195 : Ustrncmp(subject, pattern, sllen) != 0) return FAIL;
1199 /* If the local part matched, or was not being checked, check the domain using
1200 the generalized function, which supports file lookups (which may defer). The
1201 original code read as follows:
1203 return match_check_string(sdomain + 1,
1204 pdomain ? pdomain + 1 : pattern,
1205 cb->expand_setup + expand_inc, TRUE, cb->caseless, TRUE, NULL);
1207 This supported only literal domains and *.x.y patterns. In order to allow for
1208 named domain lists (so that you can right, for example, "senders=+xxxx"), it
1209 was changed to use the list scanning function. */
1211 csb.origsubject = sdomain + 1;
1212 csb.subject = cb->caseless ? string_copylc(sdomain+1) : string_copy(sdomain+1);
1213 csb.expand_setup = cb->expand_setup + expand_inc;
1214 csb.use_partial = TRUE;
1215 csb.caseless = cb->caseless;
1216 csb.at_is_special = TRUE;
1218 listptr = pdomain ? pdomain + 1 : pattern;
1219 if (valueptr) *valueptr = NULL;
1221 return match_check_list(
1222 &listptr, /* list of one item */
1223 UCHAR_MAX+1, /* impossible separator; single item */
1224 &domainlist_anchor, /* it's a domain list */
1225 &null, /* ptr to NULL means no caching */
1226 check_string, /* the function to do one test */
1227 &csb, /* its data */
1228 MCL_DOMAIN + MCL_NOEXPAND, /* domain list; don't expand */
1229 csb.subject, /* string for messages */
1230 valueptr); /* where to pass back lookup data */
1236 /*************************************************
1237 * Test whether address matches address list *
1238 *************************************************/
1240 /* This function is given an address and a list of things to match it against.
1241 The list may contain individual addresses, regular expressions, lookup
1242 specifications, and indirection via bare files. Negation is supported. The
1243 address to check can consist of just a domain, which will then match only
1244 domain items or items specified as *@domain.
1246 Domains are always lower cased before the match. Local parts are also lower
1247 cased unless "caseless" is false. The work of actually scanning the list is
1248 done by match_check_list(), with an appropriate block of arguments and a
1249 callback to check_address(). During caseless matching, it will recognize
1250 +caseful and revert to caseful matching.
1253 address address to test
1254 caseless TRUE to start in caseless state
1255 expand TRUE to allow list expansion
1256 listptr list to check against
1257 cache_bits points to cache bits for named address lists, or NULL
1258 expand_setup controls setting up of $n variables - passed through
1259 to check_address (q.v.)
1260 sep separator character for the list;
1261 may be 0 to get separator from the list;
1262 may be UCHAR_MAX+1 for one-item list
1263 valueptr where to return a lookup value, or NULL
1265 Returns: OK for a positive match, or end list after a negation;
1266 FAIL for a negative match, or end list after non-negation;
1267 DEFER if a lookup deferred
1271 match_address_list(const uschar *address, BOOL caseless, BOOL expand,
1272 const uschar **listptr, unsigned int *cache_bits, int expand_setup, int sep,
1273 const uschar **valueptr)
1275 check_address_block ab;
1276 unsigned int *local_cache_bits = cache_bits;
1279 /* RFC 2505 recommends that for spam checking, local parts should be caselessly
1280 compared. Therefore, Exim now forces the entire address into lower case here,
1281 provided that "caseless" is set. (It is FALSE for calls for matching rewriting
1282 patterns.) Otherwise just the domain is lower cases. A magic item "+caseful" in
1283 the list can be used to restore a caseful copy of the local part from the
1285 Limit the subject address size to avoid mem-exhaustion attacks. The size chosen
1286 is historical (we used to use big_buffer here). */
1288 if ((len = Ustrlen(address)) > BIG_BUFFER_SIZE) len = BIG_BUFFER_SIZE;
1289 ab.address = string_copyn(address, len);
1291 for (uschar * p = ab.address + len - 1; p >= ab.address; p--)
1293 if (!caseless && *p == '@') break;
1297 /* If expand_setup is zero, we need to set up $0 to the whole thing, in
1298 case there is a match. Can't use the built-in facilities of match_check_string
1299 (via check_address), as we may just be calling that for part of the address
1302 if (expand_setup == 0)
1304 expand_nstring[0] = string_copy(address);
1305 expand_nlength[0] = Ustrlen(address);
1309 /* Set up the data to be passed ultimately to check_address. */
1311 ab.origaddress = address;
1312 /* ab.address is above */
1313 ab.expand_setup = expand_setup;
1314 ab.caseless = caseless;
1316 return match_check_list(listptr, sep, &addresslist_anchor, &local_cache_bits,
1317 check_address, &ab, MCL_ADDRESS + (expand? 0:MCL_NOEXPAND), address,
1321 /* Simpler version of match_address_list; always caseless, expanding,
1322 no cache bits, no value-return.
1325 address address to test
1326 listptr list to check against
1327 sep separator character for the list;
1328 may be 0 to get separator from the list;
1329 may be UCHAR_MAX+1 for one-item list
1331 Returns: OK for a positive match, or end list after a negation;
1332 FAIL for a negative match, or end list after non-negation;
1333 DEFER if a lookup deferred
1337 match_address_list_basic(const uschar *address, const uschar **listptr, int sep)
1339 return match_address_list(address, TRUE, TRUE, listptr, NULL, -1, sep, NULL);
1342 /* End of match.c */